[Starlink] Re: Lost in Space: The Limits of Geolocation in a Satellite-Connected World (new article from Geoff Huston)

[J Pan <Pan@uvic.ca>]

100.64/10 is starlink's ipv4 access network. from the user side, user
router will get a unique 100.64/10 address on its wan side, and its
cgnat gateway is always fixed at 100.64.0.1. from the user 100.64/10
(100.76.147.112 in my victoria dish example) to 100.64.0.1, it's a
"long ip hop", since it's over the satellites and landing ground
station, which is under user ip traffic. for users with public ip
address on their router, 100.64/10 and 100.64.0.1 are replaced by the
corresponding public ipv4 address, respectively, but still over the
long space-ground tunnel

for ipv6, the user router has a global ipv6 address (part of its ipv6
access network) with slaac on its wan port, and a /56 prefix
delegation on its lan side for the router to distribute further

the space and ground tunnel, under the user ipv4 or ipv6 traffic,
might be ipv6 or mpls-alike, and starlink does use mpls tunnels in its
ground backbone to reach its users at a given pop
--
J Pan, UVic CSc, ECS566, 250-472-5796 (NO VM), Pan@UVic.CA, Web.UVic.CA/~pan

On Wed, Oct 1, 2025 at 9:46 AM Michael Richardson <mcr@sandelman.ca> wrote:
>
> Dr. Pan, I am still confused by the presences/need of the CGNAT layer in
> Starlink, vs your slides saying that it's "one big IPv4 hop"
>
> Is that the IPv4 100.x on the outside of the customer's access point?
> Are they unique within Starlink, or unique to each Landing Station?
> Gosh, I wish they had just done one of the v4-over-v6 mechanisms.
>
> --
> ]               Never tell me the odds!                 | ipv6 mesh networks [
> ]   Michael Richardson, Sandelman Software Works        |    IoT architect   [
> ]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [
>